SUMBER SUMBER BELAJAR

MATA KULIAH HCI (3.0)

CHAPTER 4CHAPTER 4

Semeseter Gasal (3)

Disusun olehSoedito Adjisoedarmo

DEPARTEMEN PENDIDIKAN NASIONALSTMIK WIDYA UTAMA PURWOKERTO

2013

KNOWLEDGEKNOWLEDGEREPRESENTATIONREPRESENTATION

CHAPTER 4CHAPTER 4KNOWLEDGEKNOWLEDGE

REPRESENTATIONREPRESENTATION

CHAPTER 4CHAPTER 4KNOWLEDGEKNOWLEDGE

REPRESENTATIONREPRESENTATION

REPRESENTATIONS AND KNOWLEDGETHREE REPRESENTATIONAL FAMILIES Propositional representations Analogical reprensentations Procedural representationsCONCLUSIONEXERCISES

REPRESENTATIONS AND KNOWLEDGETHREE REPRESENTATIONAL FAMILIES Propositional representations Analogical reprensentations Procedural representationsCONCLUSIONEXERCISES

Propotional representantionsSemantic feature or attributes

Semantic NetworksInheritance properties

Schemas, frames and scriptsFrame, Schema theory, Scripts and

Episodes , Plans Schemas (variables, embedded schemas,

levels of abstraction, knowledge, active process) Frames

Scripts

Analogical representationsProcedural representatations

Propotional representantionsSemantic feature or attributes

Semantic NetworksInheritance properties

Schemas, frames and scriptsFrame, Schema theory, Scripts and

Episodes , Plans Schemas (variables, embedded schemas,

levels of abstraction, knowledge, active process) Frames

Scripts

Analogical representationsProcedural representatations

RINGKASANRINGKASAN

Memory tidak hanya berisi represen-tasi fakta yang sederhana seperti yang telah dibahas di chapter 1-3. Manusia juga menyimpan penge-tahuan dari suatu events, actions and images.

Memory tidak hanya berisi represen-tasi fakta yang sederhana seperti yang telah dibahas di chapter 1-3. Manusia juga menyimpan penge-tahuan dari suatu events, actions and images.

Memory dapat dipandang sebagai suatu proses yang bekerja pada suatu representations. Pengetahuan yg dibutuhkan dan digunakan direpresen-tasikan dalam bentuk struktur penge-tahuan. Salah satu bentuk struktur pengetahuan adalah propositional networks yang menampilkan propo-sitional knowledge. Bentuk lain representasi propositional knowledge adalah : semantic features atau attributes, dan semantic networks

Memory dapat dipandang sebagai suatu proses yang bekerja pada suatu representations. Pengetahuan yg dibutuhkan dan digunakan direpresen-tasikan dalam bentuk struktur penge-tahuan. Salah satu bentuk struktur pengetahuan adalah propositional networks yang menampilkan propo-sitional knowledge. Bentuk lain representasi propositional knowledge adalah : semantic features atau attributes, dan semantic networks

Bukti dari penelitian pada mental imagenary menunjukkan bahwa beberapa bentuk dari analogical representation mungkin ada. Kenyataan image, secara fungsional mengusulkan beberapa bentuk representasi untuk informasi spatial. Meskipun demikian, dimungkinkan bahwa informasi spatial dapat direpresentasikan dengan format propositional.

Bukti dari penelitian pada mental imagenary menunjukkan bahwa beberapa bentuk dari analogical representation mungkin ada. Kenyataan image, secara fungsional mengusulkan beberapa bentuk representasi untuk informasi spatial. Meskipun demikian, dimungkinkan bahwa informasi spatial dapat direpresentasikan dengan format propositional.

Gambaran mental

Dalam mempertimbangkan know-ledge, perlu juga diperhatikan pengetahuan yang memungkinkan kita melakukan suatu aktivitas.

Contohnya, pengetahuan untuk naik sepeda, atau untuk menyentuh type writer tidak akan dipresentasikan secara sama dengan pengetahuan mengenai apa itu speda atau mesin tik. Jenis pengetahuan tersebut disebut procedural knowledge.

Dalam mempertimbangkan know-ledge, perlu juga diperhatikan pengetahuan yang memungkinkan kita melakukan suatu aktivitas.

Contohnya, pengetahuan untuk naik sepeda, atau untuk menyentuh type writer tidak akan dipresentasikan secara sama dengan pengetahuan mengenai apa itu speda atau mesin tik. Jenis pengetahuan tersebut disebut procedural knowledge.

Summary

o Memory does not solely contain representations of simple facts such as have so far been considered.

People also store knowledge of events, actions and images.

Summary

o Memory does not solely contain representations of simple facts such as have so far been considered.

People also store knowledge of events, actions and images.

o Memory can be thought of as processes operating upon represen-tations.

The knowledge people acquire and use is represented in some form of knowledge structure.

One form of knowledge structure is propositional networks which represent propositional knowledge.

Other forms of propositional knowledge representations include: semantic features or attributes, and semantic networks.

o Memory can be thought of as processes operating upon represen-tations.

The knowledge people acquire and use is represented in some form of knowledge structure.

One form of knowledge structure is propositional networks which represent propositional knowledge.

Other forms of propositional knowledge representations include: semantic features or attributes, and semantic networks.

o Evidence from experiments on mental imagery suggests that some form of analogical representation might exist.

The functional significance of images suggests some form of representation for spatial informa-tion.

However, it seems that even spatial information can be represented by a propositional format.

o Evidence from experiments on mental imagery suggests that some form of analogical representation might exist.

The functional significance of images suggests some form of representation for spatial informa-tion.

However, it seems that even spatial information can be represented by a propositional format.

o In considering knowledge, attention must also be focused on the knowledge that allows us to act.

For example, knowledge of how to ride a bicycle or how to touch-type may not be represented in the same format as our knowledge of what bicycles or typewriters are.

This is known as procedural knowledge.

o In considering knowledge, attention must also be focused on the knowledge that allows us to act.

For example, knowledge of how to ride a bicycle or how to touch-type may not be represented in the same format as our knowledge of what bicycles or typewriters are.

This is known as procedural knowledge.

REPRESENTATIONS AND KNOWLEDGE

A representation is something which stands for something else. In other words, a representation is a kind of model of something it represents.

We must assume that people construct and use models of the world in the form of knowledge representations.

REPRESENTATIONS AND KNOWLEDGE

A representation is something which stands for something else. In other words, a representation is a kind of model of something it represents.

We must assume that people construct and use models of the world in the form of knowledge representations.

REPRESENTATIONS AND KNOWLEDGE

A representation = sesuatu mewakili sesuatu yang lain

REPRESENTATIONS AND KNOWLEDGE

A representation = sesuatu mewakili sesuatu yang lain

A represent worldA represent worldA represent worldA represent worldRepresenting worldRepresenting worldRepresenting worldRepresenting world

Is the thing that is beingIs the thing that is beingrepresentedrepresentedIs the thing that is beingIs the thing that is beingrepresentedrepresented

Whatever is doing therepresentingWhatever is doing therepresenting

ArsitekArsitek

Perencanaan membuat (rancangan) rumahadalah suatu representasibeberapa aspek dari rumah

Perencanaan membuat (rancangan) rumahadalah suatu representasibeberapa aspek dari rumah

Rumah adalah sesuatu yang direpresentasikan (is being represented)Rumah adalah sesuatu yang direpresentasikan (is being represented)

Perencanaan adalah sesuatu yang melakukan representasiPerencanaan adalah sesuatu yang melakukan representasi

We can distinguish between a represented world and a represen-ting world.

The represented world is the thing that is being represented, and the representing world is whatever is doing the representing.

We can distinguish between a represented world and a represen-ting world.

The represented world is the thing that is being represented, and the representing world is whatever is doing the representing.

For example, an architect's plan of a house is a representation of some aspect of the house.

The house is what is being represented and the plan is what is doing the representing. Palmer (1978) identified five features for a representational system as follows:

For example, an architect's plan of a house is a representation of some aspect of the house.

The house is what is being represented and the plan is what is doing the representing. Palmer (1978) identified five features for a representational system as follows:

1 What is the represented world.2 What is the representing world.3 What aspects of the represented world are being modelled.4 What aspects of the representing world are doing the modelling.5 What are the correspondences between the two worlds.

1 What is the represented world.2 What is the representing world.3 What aspects of the represented world are being modelled.4 What aspects of the representing world are doing the modelling.5 What are the correspondences between the two worlds.

Five features for a representational systemFive features for a representational system

1. Apa yang direpresentasikan2. Apa yang merepresantikan 3. Aspek apa dari yang direpresetasikan dimodelkan4. Aspek apa dari yang direpresentasikan yang mengerjakan model 5. Yang berhubungan antara 1-2, 3-4

1. Apa yang direpresentasikan2. Apa yang merepresantikan 3. Aspek apa dari yang direpresetasikan dimodelkan4. Aspek apa dari yang direpresentasikan yang mengerjakan model 5. Yang berhubungan antara 1-2, 3-4

Represented Representing worldWorld

Objects a A 15 7 h B 13 9

PropertiesHeight

Not directly Numeric Numericrepresented

Relation a taller than b TALLER GREATER LESSTHAN (A,B) THAN THAN

Figure 4.1 Different representations of the property of height

Represented Representing worldWorld

Objects a A 15 7 h B 13 9

PropertiesHeight

Not directly Numeric Numericrepresented

Relation a taller than b TALLER GREATER LESSTHAN (A,B) THAN THAN

Figure 4.1 Different representations of the property of height

Each form of representation captures someaspects of the relationship between the two objectEach form of representation captures someaspects of the relationship between the two object

By way of example, consider how in Fig. 4.1 the property of height' might be represented in a number of ways.

Each form of representation captures some aspect of the relationship between the two objects.

It should be noted that a sixth feature of a representational system is the purpose for which it was constructed.

By way of example, consider how in Fig. 4.1 the property of height' might be represented in a number of ways.

Each form of representation captures some aspect of the relationship between the two objects.

It should be noted that a sixth feature of a representational system is the purpose for which it was constructed.

One can distinguish between a person's mental representation of the world and a theory or model of a person's representation of the world (perhaps as would be constructed by a psychologist or an expert system).

Psychological theories of knowledge are, in fact, models of people's represented knowledge of the world.

Psychological theories of knowledge representation are not concerned with modelling the phenomena in the world.

One can distinguish between a person's mental representation of the world and a theory or model of a person's representation of the world (perhaps as would be constructed by a psychologist or an expert system).

Psychological theories of knowledge are, in fact, models of people's represented knowledge of the world.

Psychological theories of knowledge representation are not concerned with modelling the phenomena in the world.

They are concerned with modelling the knowledge people have of the world and how that knowledge is organized and structured.

While a person's represented knowledge of the world is not necessarily identical to the objects of that world, their knowledge may reflect much of the structuring of the world and, in many cases, imposes a structure on the world.

They are concerned with modelling the knowledge people have of the world and how that knowledge is organized and structured.

While a person's represented knowledge of the world is not necessarily identical to the objects of that world, their knowledge may reflect much of the structuring of the world and, in many cases, imposes a structure on the world.

THREE REPRESENTATIONAL FAMILIES

Three main families of knowledge representation can be identified from the many theories and models of the structuring of human knowledge that abound.

THREE REPRESENTATIONAL FAMILIES

Three main families of knowledge representation can be identified from the many theories and models of the structuring of human knowledge that abound.

THREE REPRESENTATIONAL FAMILIES

1 Propositional representations

This is the most widely discussed family of knowledge representational formats and includes a wide range of theories and models.

These theories have in common the fact that they represent knowledge as a set of discrete symbols or propositions, concepts, objects and features, and relations can all be represented by propositional representations.

THREE REPRESENTATIONAL FAMILIES

1 Propositional representations

This is the most widely discussed family of knowledge representational formats and includes a wide range of theories and models.

These theories have in common the fact that they represent knowledge as a set of discrete symbols or propositions, concepts, objects and features, and relations can all be represented by propositional representations.

a set of discrete symbols or a set of discrete symbols or propositions, propositions,

concepts, concepts, objects and features, objects and features,

and relations can all be represented and relations can all be represented by propositional representations.by propositional representations.

knowledgeknowledge

diskrit

Propositional representations of Propositional representations of

Logical represtationMap, picture

continues

2 Analogical representations

This form of representation has been postulated to explain phenomena such as mental images.

Analogical representations maintain a close correspondence between the representing and represented world.

In this form of representation the variable parameters of the representation are assumed to be continuous, in the same way that voltages, maps and pictures all have continous properties.

2 Analogical representations

This form of representation has been postulated to explain phenomena such as mental images.

Analogical representations maintain a close correspondence between the representing and represented world.

In this form of representation the variable parameters of the representation are assumed to be continuous, in the same way that voltages, maps and pictures all have continous properties.

3. Procedural representations

This is perhaps the least well-developed class or family of theories of knowledge representation.

Theories within this family are postulated to explain how a person's knowledge of actions are represented. For example, the knowledge a person possesses that enables them to walk, talk, add two numbers together, play chess, ride a bicycle, etc., must be represented in some format.

3. Procedural representations

This is perhaps the least well-developed class or family of theories of knowledge representation.

Theories within this family are postulated to explain how a person's knowledge of actions are represented. For example, the knowledge a person possesses that enables them to walk, talk, add two numbers together, play chess, ride a bicycle, etc., must be represented in some format.

The knowledge used for executing actions is assumed to be represented as procedural knowledge.

Procedural knowledge is directly interpretable by an action system.

For example, the knowledge one might have of riding a bicycle is assumed to be directly tied up with the activity of riding a bicycle and can only be accessed by carrying out the activity (i.e. riding the bicycle).

The knowledge used for executing actions is assumed to be represented as procedural knowledge.

Procedural knowledge is directly interpretable by an action system.

For example, the knowledge one might have of riding a bicycle is assumed to be directly tied up with the activity of riding a bicycle and can only be accessed by carrying out the activity (i.e. riding the bicycle).

It would be wrong to think that people's knowledge can be discretely segmented into different forms of representation or

that the types of knowledge considered by each of these three forms of representation were in some way mutually exclusive.

It would be wrong to think that people's knowledge can be discretely segmented into different forms of representation or

that the types of knowledge considered by each of these three forms of representation were in some way mutually exclusive.

For example, the knowledge a person has about a particular computer application would include knowledge of what it could be used for, what particular windows or menus looked like, and what-actions were required to position the cursor using the mouse.

For example, the knowledge a person has about a particular computer application would include knowledge of what it could be used for, what particular windows or menus looked like, and what-actions were required to position the cursor using the mouse.

Moreover, it would be possible to visualize the shape of the cursor to describe how to hold the mouse and to carry out the actions of moving the cursor directly to a target.

People's knowledge of the world is rich and polymorphic.

Moreover, it would be possible to visualize the shape of the cursor to describe how to hold the mouse and to carry out the actions of moving the cursor directly to a target.

People's knowledge of the world is rich and polymorphic.

The types of knowledge representation referred to above and discussed in further detail below are best thought of

as contributing factors to a full understanding of the different functional attributes of knowledge.

The types of knowledge representation referred to above and discussed in further detail below are best thought of

as contributing factors to a full understanding of the different functional attributes of knowledge.

Before considering each of these

families of knowledge representation in

turn, it is worth remembering that

representations require processes that

can interpret or use the

representational structure.

Before considering each of these

families of knowledge representation in

turn, it is worth remembering that

representations require processes that

can interpret or use the

representational structure.

PropositionalLogicalProcedural

Referring back to Fig. 4.1,. before any statement or judgement could be made about the property height', there would

need to be a process for evaluating the expression TALLER THAN (A,B), for comparing the two line lengths or for calculating the difference between the numeric values.

Referring back to Fig. 4.1,. before any statement or judgement could be made about the property height', there would

need to be a process for evaluating the expression TALLER THAN (A,B), for comparing the two line lengths or for calculating the difference between the numeric values.

Propositional representations

This form of representation has been given the most 'attention in cognitive psychology.’

It assumes that knowledge is represented by a collection of symbols, and includes a wide range of theories that attempt to describe how knowledge is represented for an even wider range of phenomena.

Propositional representations

This form of representation has been given the most 'attention in cognitive psychology.’

It assumes that knowledge is represented by a collection of symbols, and includes a wide range of theories that attempt to describe how knowledge is represented for an even wider range of phenomena.

Semantic features or attributes

Concepts can be represented by sets of semantic features or attributes.

This form of representation has borrowed much from set theory and in particular questions the relationships between sets of features.

A concept is thought to be represented by a set of weighted features which can then be considered in the following terms:

Semantic features or attributes

Concepts can be represented by sets of semantic features or attributes.

This form of representation has borrowed much from set theory and in particular questions the relationships between sets of features.

A concept is thought to be represented by a set of weighted features which can then be considered in the following terms:

Disjoint Non-overlapping attributesOverlap Some but not all attributes in commonNested All of X are in YIdentical Exact same features in X as in Y

Disjoint Non-overlapping attributesOverlap Some but not all attributes in commonNested All of X are in YIdentical Exact same features in X as in Y

The attributes or features of a concept can have differing weights to represent the salience and importance of the particular attribute or feature.

However, it is particularly continous as to what constitutes a salient or important feature or attribute of a concept. It is also very difficult to say how importance or salience can be characterized or measured.

The attributes or features of a concept can have differing weights to represent the salience and importance of the particular attribute or feature.

However, it is particularly continous as to what constitutes a salient or important feature or attribute of a concept. It is also very difficult to say how importance or salience can be characterized or measured.

For example, what are the salient features of a program that allows you to recognize a piece of text is not a program?

The language in which a program is written is clearly a feature of the program, but something is not a program because it is written in, say Pascal or Lisp.

For example, what are the salient features of a program that allows you to recognize a piece of text is not a program?

The language in which a program is written is clearly a feature of the program, but something is not a program because it is written in, say Pascal or Lisp.

Bhs pemrograman-- featureTidak semua yang ditulis program

Many things that are written in programming languages are definitely not programs.

Conversely, there are many pro-grams that are not written in conventional programming languages that are indeed programs (e.g., knitting patterns are programmed instructions for a person to knit a garment).

Many things that are written in programming languages are definitely not programs.

Conversely, there are many pro-grams that are not written in conventional programming languages that are indeed programs (e.g., knitting patterns are programmed instructions for a person to knit a garment).

Banyak yg ditulis dlm bhs program bukan programBanyak program yang tidak ditulis dalm bhs program

Early experiments carried out to investigate the features of conceptual knowledge;

and the psychological structure of word meaning led to the development of initial theories of propositional knowledge being represented in terms of semantic features or attributes.

Early experiments carried out to investigate the features of conceptual knowledge;

and the psychological structure of word meaning led to the development of initial theories of propositional knowledge being represented in terms of semantic features or attributes.

THE FEATURES OF CONCEPTUAL KNOWLEDGE

psychological structure of word meaning

led to the development of initial theories of propositional knowledge

represented in terms of semantic features or attributes.

A classic experiment was that of Collins and Quillian (1969) who found that subjects in their experiments took longer to judge a statement such as 'canary eats food' as being true than a statement such as 'canary has feathers', which in turn took longer to judge than 'canary is yellow'.

A classic experiment was that of Collins and Quillian (1969) who found that subjects in their experiments took longer to judge a statement such as 'canary eats food' as being true than a statement such as 'canary has feathers', which in turn took longer to judge than 'canary is yellow'.

From these results it was suggested that information is stored hierarchically, and that properties specific to the concept canary' are stored directly with that concept.

However, properties specific to the concept 'bird' would be stored with the 'bird' concept and not the 'canary' concept, and properties specific to the concept 'animal' would be stored with the 'animal' concept not the 'bird' concept (Fig. 4.2). Furthermore, the further up the hierarchy to be searched, the longer time it takes.

From these results it was suggested that information is stored hierarchically, and that properties specific to the concept canary' are stored directly with that concept.

However, properties specific to the concept 'bird' would be stored with the 'bird' concept and not the 'canary' concept, and properties specific to the concept 'animal' would be stored with the 'animal' concept not the 'bird' concept (Fig. 4.2). Furthermore, the further up the hierarchy to be searched, the longer time it takes.

Animal ( eats food, has offspring, breaths oxygen)Bird ((is an animal) (has feathers, can fly, lays eggs))Canary ((is a bird) (has yellow feathers, lives in exotic places))

Figure 4.2 An example hierarchy for the concepts 'animal', 'bird' and 'canary'

Animal ( eats food, has offspring, breaths oxygen)Bird ((is an animal) (has feathers, can fly, lays eggs))Canary ((is a bird) (has yellow feathers, lives in exotic places))

Figure 4.2 An example hierarchy for the concepts 'animal', 'bird' and 'canary'

However, further experiments by Smith, Shoben and Rips (1974) found some exceptions to the experimental results of Collins and Quillian (op. cit.).

In a similar study to that of Collins and Quillian.

However, further experiments by Smith, Shoben and Rips (1974) found some exceptions to the experimental results of Collins and Quillian (op. cit.).

In a similar study to that of Collins and Quillian.

Smith et al. found that subjects were able to judge sentences such as 'a robin is a bird' as being true, faster than they were able to judge sentences such as 'a chicken is a bird'.

Smith et al. found that subjects were able to judge sentences such as 'a robin is a bird' as being true, faster than they were able to judge sentences such as 'a chicken is a bird'.

Results such as these led Smith et al. to propose that the more typical an instance is of a category, the more quickly it can be verified as belonging to that category.

A two stage process for identifying category membership was proposed as follows:

Results such as these led Smith et al. to propose that the more typical an instance is of a category, the more quickly it can be verified as belonging to that category.

A two stage process for identifying category membership was proposed as follows:

1. A quick comparison of all features is made:

if the comparison is relatively good answer yes, if poor answer no, if undecided then:

2. A more elaborate comparison process is applied to identify the defining features.

1. A quick comparison of all features is made:

if the comparison is relatively good answer yes, if poor answer no, if undecided then:

2. A more elaborate comparison process is applied to identify the defining features.

A two stage process for identifying category membership was proposed as follows:

Therefore, judgements on the truth

value of statements of the sort,

'sparrow is a bird' would be quick, as

would be 'door is a bird', but 'bat is a

bird' would take longer to judge.

Therefore, judgements on the truth

value of statements of the sort,

'sparrow is a bird' would be quick, as

would be 'door is a bird', but 'bat is a

bird' would take longer to judge.

Thus, this type of representation assumes

that conceptual knowledge is represented

by a set of features and that those

features include defining attributes of the

concept being represented as well as

attributes that are characteristic of the

concept.

Thus, this type of representation assumes

that conceptual knowledge is represented

by a set of features and that those

features include defining attributes of the

concept being represented as well as

attributes that are characteristic of the

concept.

DefiningCharacteristic

Defining attributes are those which identify the concept as being an instance of a type, for example, the attributes of a 'sparrow' that make it identifiable as being an instance of the 'bird' category.

A characteristic attribute is one which the particular instance has but which does not provide sufficient definition for it to be included or excluded from the class.

Defining attributes are those which identify the concept as being an instance of a type, for example, the attributes of a 'sparrow' that make it identifiable as being an instance of the 'bird' category.

A characteristic attribute is one which the particular instance has but which does not provide sufficient definition for it to be included or excluded from the class.

ConceptConcept(conseptual(conseptualknowledge) knowledge)

Defining attributes(=c is instance of type)

Characteristic attributes(yg dimiliki particular instance)

Represen

ted by

Semantic representation

Effective model Propositionalreprentation

Semantic representations such as the feature-attribute set are effective models of the class of experimental data produced from studies such as those of Collins and Quillian; however, these models do have the following limitations:

Semantic representations such as the feature-attribute set are effective models of the class of experimental data produced from studies such as those of Collins and Quillian; however, these models do have the following limitations:

1. All the work is with simple, nominal concepts such as 'animals' or 'fruit'. It is not clear how, for example, to represent simple facts or events such as 'songs’ are for 'singing' or 'John ran away'.

2. The representations cannot handle distinctions such as, 'sparrow is a bird', ‘robin is a bird', but 'sparrow is not a robin'.

1. All the work is with simple, nominal concepts such as 'animals' or 'fruit'. It is not clear how, for example, to represent simple facts or events such as 'songs’ are for 'singing' or 'John ran away'.

2. The representations cannot handle distinctions such as, 'sparrow is a bird', ‘robin is a bird', but 'sparrow is not a robin'.

3. Quantification is a further problem

For example, the meaning of the sentence, 'everyone kissed someone' is different from the meaning of 'someone was kissed by everyone'.

3. Quantification is a further problem

For example, the meaning of the sentence, 'everyone kissed someone' is different from the meaning of 'someone was kissed by everyone'.

Semantic Networks

Semantic networks have been, and still are, used to represent the associations that exist between conceptual knowledge in memory (see, for example, Quillian, 1966).

Knowledge is represented by a kind of directed labelled graph with nodes interrelated by relations (Fig. 4.3).

Semantic Networks

Semantic networks have been, and still are, used to represent the associations that exist between conceptual knowledge in memory (see, for example, Quillian, 1966).

Knowledge is represented by a kind of directed labelled graph with nodes interrelated by relations (Fig. 4.3).

Nodes represent concepts in memory; relations are associations among nodes; and relations are labelled and directed.

The meaning of a concept (or node) is given by the pattern of its 'relations among which it participates’.

Nodes represent concepts in memory; relations are associations among nodes; and relations are labelled and directed.

The meaning of a concept (or node) is given by the pattern of its 'relations among which it participates’.

is a

Dog Boxer

Likes Chewed Chewed

Meat stole Siamse

Figure 4.3 A simple Semantic network

{Animal} {Person}eats food subset animalbreathes air has as part legshas mass has as part armsHas as part limbs

Figure 4.4 Example of the inheritance properties between the two concepts, 'animal' and ‘person’

{Animal} {Person}eats food subset animalbreathes air has as part legshas mass has as part armsHas as part limbs

Figure 4.4 Example of the inheritance properties between the two concepts, 'animal' and ‘person’

Inheritance properties

One attribute of the semantic network formalism is the convenience with which the property of inheritance is formulated. Inheritance is the notion that certain features may be shared between concepts.

It is dependent on the existence of categories and the hierarchical structure of those categories.

Inheritance properties

One attribute of the semantic network formalism is the convenience with which the property of inheritance is formulated. Inheritance is the notion that certain features may be shared between concepts.

It is dependent on the existence of categories and the hierarchical structure of those categories.

It assumes that child members of the category inherit some of the features from the parent members of the category.

For example, the concept 'house' might be categorized as a member of the supracategory 'building'. Buildings might be considered to have features that include 'walls', 'floor', and 'roof'.

It assumes that child members of the category inherit some of the features from the parent members of the category.

For example, the concept 'house' might be categorized as a member of the supracategory 'building'. Buildings might be considered to have features that include 'walls', 'floor', and 'roof'.

Buildings might be considered to have features that include 'walls', 'floor', and 'roof'.

The concept house might also be expected to have the features of 'walls', 'floor, 'roof'~ in which case we might assume that houses inherit these features from the supracategory 'building'. However, is it the case that all members of the category inherit all features from the supracategory?

Buildings might be considered to have features that include 'walls', 'floor', and 'roof'.

The concept house might also be expected to have the features of 'walls', 'floor, 'roof'~ in which case we might assume that houses inherit these features from the supracategory 'building'. However, is it the case that all members of the category inherit all features from the supracategory?

In Fig. 4.4 the concept ‘animal' has an is a relational link from the concept person'. One feature of the concept 'animal' is that of eats food. Where eats is a slot and food is a filler of that slot. The concept 'person' inherits all the features of the concept 'animal', but has its own specified slot-filler has aspart legs and arms. In order for the semantic network to function there must exist a basic set of nodes and relations.

In Fig. 4.4 the concept ‘animal' has an is a relational link from the concept person'. One feature of the concept 'animal' is that of eats food. Where eats is a slot and food is a filler of that slot. The concept 'person' inherits all the features of the concept 'animal', but has its own specified slot-filler has aspart legs and arms. In order for the semantic network to function there must exist a basic set of nodes and relations.

One important class of relations is the type indicating that one node is an instance of the class pointed to by the relation. Two important type relations are is a and Subset. Thus. instances and subsets inherit the properties of their types. For example. 'birds' have feathers and can fly: by inheritance this applies to all birds.However. 'ostriches' are an instance of the class 'bird' but they cannot fly.

One important class of relations is the type indicating that one node is an instance of the class pointed to by the relation. Two important type relations are is a and Subset. Thus. instances and subsets inherit the properties of their types. For example. 'birds' have feathers and can fly: by inheritance this applies to all birds.However. 'ostriches' are an instance of the class 'bird' but they cannot fly.

The solution to this problem has been to add another node to the 'ostrich' definition which specifies that it cannot fly. Thus exceptions to the inherited class properties are important defining features of an instance. This appears to be inconsistent with the view that instances inherit all the properties of the class. Consequently, we cannot assume that concepts are represented in simple class hierarchies with full class inheritance. Process rules for determining and identifying an instance have been formulated as follows:

The solution to this problem has been to add another node to the 'ostrich' definition which specifies that it cannot fly. Thus exceptions to the inherited class properties are important defining features of an instance. This appears to be inconsistent with the view that instances inherit all the properties of the class. Consequently, we cannot assume that concepts are represented in simple class hierarchies with full class inheritance. Process rules for determining and identifying an instance have been formulated as follows:

1 Look first at the node of the concept.2 If the information is not found, go up one

node along the type relation and apply the property of inheritance.

3 Repeat 2 until there is a success or no more nodes.

This will always find the lowest (most specific) level of the relationship that applies. However, it will never notice inconsistencies

1 Look first at the node of the concept.2 If the information is not found, go up one

node along the type relation and apply the property of inheritance.

3 Repeat 2 until there is a success or no more nodes.

This will always find the lowest (most specific) level of the relationship that applies. However, it will never notice inconsistencies

Schemas, frames and scripts

Semantic features and semantic networks focus on basic, elementary units of knowledge.

The semantic-features approach focuses on the representation of word meaning. Semantic networks include lexical- and sentential-level knowledge. But neither approach represents higher level knowledge structures (suprasentential level), such as events in stories.

Schemas, frames and scripts

Semantic features and semantic networks focus on basic, elementary units of knowledge.

The semantic-features approach focuses on the representation of word meaning. Semantic networks include lexical- and sentential-level knowledge. But neither approach represents higher level knowledge structures (suprasentential level), such as events in stories.

Four strands of theoretical and empirical research have begun to address this type of knowledge representation. These are as follows:

o Frames (Minsky) -o Schema theory (Rumelhart)o Scripts and episodes (Schank)o Plans (Abelson)

Four strands of theoretical and empirical research have begun to address this type of knowledge representation. These are as follows:

o Frames (Minsky) -o Schema theory (Rumelhart)o Scripts and episodes (Schank)o Plans (Abelson)

Both episodes and plans have to some degree been combined in script theory (Schank and Abelson, 1977).

These theories try to structure knowledge into higher order representational units. They attempt to add structure to model the higher level relationships between the lower order units (i.e. concept features and attributes).

Both episodes and plans have to some degree been combined in script theory (Schank and Abelson, 1977).

These theories try to structure knowledge into higher order representational units. They attempt to add structure to model the higher level relationships between the lower order units (i.e. concept features and attributes).

Schemas

Schemas are data structures for representing the generic concepts stored in memory.

There are assumed to be schemas for generalized concepts underlying objects, situations, events, sequen-ces of events, actions and sequences of actions.

Schemas can be thought of as producing models of the world.

Schemas

Schemas are data structures for representing the generic concepts stored in memory.

There are assumed to be schemas for generalized concepts underlying objects, situations, events, sequen-ces of events, actions and sequences of actions.

Schemas can be thought of as producing models of the world.

To process information with the use of a schema is to generate one or more models and then determine which of the generated models best fits the incoming information.

A particular configuration of schemas which are used to construct the best-fit model constitutes an interpretation of the world.

To process information with the use of a schema is to generate one or more models and then determine which of the generated models best fits the incoming information.

A particular configuration of schemas which are used to construct the best-fit model constitutes an interpretation of the world.

o They have variables.o They can be embedded in other schemas.o They represent knowledge at all levels of abstraction.o They represent knowledge rather than definitions.o They are active recognition devices

whose processing is aimed at the evaluation of their goodness of fit to the data being processed.

o They have variables.o They can be embedded in other schemas.o They represent knowledge at all levels of abstraction.o They represent knowledge rather than definitions.o They are active recognition devices

whose processing is aimed at the evaluation of their goodness of fit to the data being processed.

Schemas have a number of characteristics as follows:Schemas have a number of characteristics as follows:

Variables (karakteristik Schema)

Schemas for any concept contain a fixed part for those characteristics which are most true of exemplars of a concept and a variable part. For example:

Fixed VariableDog has legs brown

has tail 30 ins high

Therefore legs and tail are assumed to be fixed or constant while the colour and size are assumed to be variable features of the concept dog'. Variables can have default values; for example, the default value for the colour of dogs may be brown. Consider the following story extract:

John saw the balloon man coming down the street. He remembered his brother's birthday and rushed into the house.

Variables (karakteristik Schema)

Schemas for any concept contain a fixed part for those characteristics which are most true of exemplars of a concept and a variable part. For example:

Fixed VariableDog has legs brown

has tail 30 ins high

Therefore legs and tail are assumed to be fixed or constant while the colour and size are assumed to be variable features of the concept dog'. Variables can have default values; for example, the default value for the colour of dogs may be brown. Consider the following story extract:

John saw the balloon man coming down the street. He remembered his brother's birthday and rushed into the house.

To interpret this story there might be schemas for tradesmen passing through a residential community selling toys to children. One schema might be characterized as follows:

Fixed Relationships between characters of the tradesman-customer drama

Variable Particular individuals filling specific roles

Thus : John = Buyer (variable)Assume: John is a young boy (although not specified in story)

Thus: Default value of Age, Buyer = Childhood.

To interpret this story there might be schemas for tradesmen passing through a residential community selling toys to children. One schema might be characterized as follows:

Fixed Relationships between characters of the tradesman-customer drama

Variable Particular individuals filling specific roles

Thus : John = Buyer (variable)Assume: John is a young boy (although not specified in story)

Thus: Default value of Age, Buyer = Childhood.

In the above example, assumptions about the age of John have been made to make the interpretation of the story fit with an existing schema. However, later in the story we try be informed that John is, in fact, in his twenties and that his brother is in his thirties, and that he is a keen hot-air balloonist. Therefore, the schema that was applied may have been completely wrong, in that the sequence of events was not concerned with tradesmen selling balloons to young children, but instead with hot-air balloonists preparing for a day's flight on the occasion of the birthday of one of their members.

In the above example, assumptions about the age of John have been made to make the interpretation of the story fit with an existing schema. However, later in the story we try be informed that John is, in fact, in his twenties and that his brother is in his thirties, and that he is a keen hot-air balloonist. Therefore, the schema that was applied may have been completely wrong, in that the sequence of events was not concerned with tradesmen selling balloons to young children, but instead with hot-air balloonists preparing for a day's flight on the occasion of the birthday of one of their members.

Embedded schemas (karakteristik Schema)

A schema consists of a configuration of subschemas. However, some schemas are primitive, that is, not decomposable.

For example, a schema for recognizing the human body might be made up of the following subschemas:

HUMANBODY (HEAD, TRUNK, LIMBS)HEAD (FACE, EARS, HAIR)FACE (2 EYES, NOSE, MOUTH)EYE (IRIS, UPPER-LID, LOWER-LID)

Embedded schemas (karakteristik Schema)

A schema consists of a configuration of subschemas. However, some schemas are primitive, that is, not decomposable.

For example, a schema for recognizing the human body might be made up of the following subschemas:

HUMANBODY (HEAD, TRUNK, LIMBS)HEAD (FACE, EARS, HAIR)FACE (2 EYES, NOSE, MOUTH)EYE (IRIS, UPPER-LID, LOWER-LID)

Thus schemas propose a hierarchy of levels rather than a single level.

In the example above, the components of the embedded HUMAN BODY schema have primitive subschemas for IRIS, UPPER-LID and LOWER-LID.

Thus schemas propose a hierarchy of levels rather than a single level.

In the example above, the components of the embedded HUMAN BODY schema have primitive subschemas for IRIS, UPPER-LID and LOWER-LID.

Levels of abstraction(karakteristik Schema)

Schemas represent knowledge at all levels of abstraction, for instance, ideological, sentences, words, form of letters, etc. Schema theories assume that the human memory system contains countless packets' of knowledge. Each packet specifies a configuration of other packets (or subschemas) and each packet may vary in complexity and level of application.

Levels of abstraction(karakteristik Schema)

Schemas represent knowledge at all levels of abstraction, for instance, ideological, sentences, words, form of letters, etc. Schema theories assume that the human memory system contains countless packets' of knowledge. Each packet specifies a configuration of other packets (or subschemas) and each packet may vary in complexity and level of application.

Knowledge (karakteristik Schema)

All our knowledge is assumed to be embedded in schema. Knowledge can be thought of in terms of semantic and episodic components. Semantic components include dictionary knowledge (i.e. the essential aspects of word meaning), and encyclopaedic knowledge (i.e. facts and relationships). Episodic knowledge focuses on experiential components; for example, that you once fed a blackbird with breadcrumbs all winter. Schemas are assumed to exist for both semantic and episodic components of knowledge.

Knowledge (karakteristik Schema)

All our knowledge is assumed to be embedded in schema. Knowledge can be thought of in terms of semantic and episodic components. Semantic components include dictionary knowledge (i.e. the essential aspects of word meaning), and encyclopaedic knowledge (i.e. facts and relationships). Episodic knowledge focuses on experiential components; for example, that you once fed a blackbird with breadcrumbs all winter. Schemas are assumed to exist for both semantic and episodic components of knowledge.

Active processes Active processes

Each schema is assumed to have Each schema is assumed to have a process. The processing carried a process. The processing carried out by a schema includes out by a schema includes evaluating its goodness of fit to evaluating its goodness of fit to the incoming information, binding the incoming information, binding its variables and sending its variables and sending messages to other schemas. messages to other schemas. There are two types of data-There are two types of data-sources; bottom-up and top-down:sources; bottom-up and top-down:

Active processes Active processes

Each schema is assumed to have Each schema is assumed to have a process. The processing carried a process. The processing carried out by a schema includes out by a schema includes evaluating its goodness of fit to evaluating its goodness of fit to the incoming information, binding the incoming information, binding its variables and sending its variables and sending messages to other schemas. messages to other schemas. There are two types of data-There are two types of data-sources; bottom-up and top-down:sources; bottom-up and top-down:

Active processesActive processes

Each schema is assumed to have Each schema is assumed to have a process. The processing carried a process. The processing carried out by a schema includes out by a schema includes evaluating its goodness of fit to evaluating its goodness of fit to the incoming information, binding the incoming information, binding its variables and sending its variables and sending messages to other schemas. messages to other schemas. There are two types of data-There are two types of data-sources; bottom-up and top-down:sources; bottom-up and top-down:

Active processesActive processes

Each schema is assumed to have Each schema is assumed to have a process. The processing carried a process. The processing carried out by a schema includes out by a schema includes evaluating its goodness of fit to evaluating its goodness of fit to the incoming information, binding the incoming information, binding its variables and sending its variables and sending messages to other schemas. messages to other schemas. There are two types of data-There are two types of data-sources; bottom-up and top-down:sources; bottom-up and top-down:

Active processes (karakteristik Schema)

Each schema is assumed to have a process. The processing carried out by a schema includes evaluating its goodness of fit to the incoming information, binding its variables and sending messages to other schemas. There are two types of data-sources; bottom-up and top-down:

Active processes (karakteristik Schema)

Each schema is assumed to have a process. The processing carried out by a schema includes evaluating its goodness of fit to the incoming information, binding its variables and sending messages to other schemas. There are two types of data-sources; bottom-up and top-down:

o Bottom-up provides information from subschemas about how well they fit the input.

o Top-down provides information from supraschemas about the degree of cer-tainty of their relevance to structuring the input.

Interpretation of meaning consists of top-down and bottom-up processing in repeated loops. The set of schemas that have the best goodness-of-fit constitutes an interpretation.

o Bottom-up provides information from subschemas about how well they fit the input.

o Top-down provides information from supraschemas about the degree of cer-tainty of their relevance to structuring the input.

Interpretation of meaning consists of top-down and bottom-up processing in repeated loops. The set of schemas that have the best goodness-of-fit constitutes an interpretation.

FramesFrames are very similar to schemas in that they provide variable slots which can take the particular fillers for an instantiated frame. A frame is instantiated when it is provided with the particular details for a given context. Frames can exist at a number of different levels, with high-level or generalized frame structures and also low-level specific frame structures in the same way that schemas can be embedded.

FramesFrames are very similar to schemas in that they provide variable slots which can take the particular fillers for an instantiated frame. A frame is instantiated when it is provided with the particular details for a given context. Frames can exist at a number of different levels, with high-level or generalized frame structures and also low-level specific frame structures in the same way that schemas can be embedded.

The higher level frames call up the lower level frames which then detail how some component of the higher level frame is to be further interpreted.

Unlike schema, frames do not possess active processors which interpret the world; instead, frame theory is restricted to addressing the representation of knowledge and ignores any kind of processing that might be carried out on the knowledge.

The higher level frames call up the lower level frames which then detail how some component of the higher level frame is to be further interpreted.

Unlike schema, frames do not possess active processors which interpret the world; instead, frame theory is restricted to addressing the representation of knowledge and ignores any kind of processing that might be carried out on the knowledge.

Recently, frame-based representations have been used to model the knowledge people possess about a task domain (Keane and Johnson, 1987).

In the context of human-computer interaction, Keane and Johnson carried out an empirical analysis of the knowledge people possessed about various tasks such as ‘arranging a meeting'.

Recently, frame-based representations have been used to model the knowledge people possess about a task domain (Keane and Johnson, 1987).

In the context of human-computer interaction, Keane and Johnson carried out an empirical analysis of the knowledge people possessed about various tasks such as ‘arranging a meeting'.

The analysis identified knowledge in terms of goals, plans, macro-actions (procedures), micro-actions, and objects. The goals were represented as a general goal frame which then became particularized into subgoals at lower level frames.

Each frame included either a goal or subgoal depending upon the level of the frame. Within the frame a plan for carrying out the task was detailed. This plan became more specialized as frames were rewritten into lower level frames.

The analysis identified knowledge in terms of goals, plans, macro-actions (procedures), micro-actions, and objects. The goals were represented as a general goal frame which then became particularized into subgoals at lower level frames.

Each frame included either a goal or subgoal depending upon the level of the frame. Within the frame a plan for carrying out the task was detailed. This plan became more specialized as frames were rewritten into lower level frames.

Task: Arrange a project meetingPlan (meeting (project)) Consult (information source, information token, project meeting) ldentify (information source, information token, project meeting) Search (information source, information token) Retrieve (information token, information source) Store (information token, project meeting, working memory)Select (media message) Identify (long-term memory, constraints, project meeting) Choose (media, constraints)Send message (meeting, (ploject)) Consult (information token, information source, letter) Identify (information source, information token, letter) Search (information source, information token) Retrieve (information token, information source) Store (information token, letter, working memory)

Task: Arrange a project meetingPlan (meeting (project)) Consult (information source, information token, project meeting) ldentify (information source, information token, project meeting) Search (information source, information token) Retrieve (information token, information source) Store (information token, project meeting, working memory)Select (media message) Identify (long-term memory, constraints, project meeting) Choose (media, constraints)Send message (meeting, (ploject)) Consult (information token, information source, letter) Identify (information source, information token, letter) Search (information source, information token) Retrieve (information token, information source) Store (information token, letter, working memory)

Represent (information token, message) Write (information token, message, media) Compare (message, information token) Edit (information token, message) Store (message copy, message file, media) Execute (transaction requirements, message)

Figure 4.5 An example of a frame-based representation of the knowledge used to create and send a message to arrange a project meeting (from Keane and Johnson, 1987)

Represent (information token, message) Write (information token, message, media) Compare (message, information token) Edit (information token, message) Store (message copy, message file, media) Execute (transaction requirements, message)

Figure 4.5 An example of a frame-based representation of the knowledge used to create and send a message to arrange a project meeting (from Keane and Johnson, 1987)

The procedures used in the task were represented by the groupings of actions within a particular frame.

The individual activities in the task were represented as procedures and actions (or macro- and micro-actions) with procedures being high-level activities and actions being low-level activities. The object sets were then the entities on which the activities were performed or otherwise associated with the activities.

The procedures used in the task were represented by the groupings of actions within a particular frame.

The individual activities in the task were represented as procedures and actions (or macro- and micro-actions) with procedures being high-level activities and actions being low-level activities. The object sets were then the entities on which the activities were performed or otherwise associated with the activities.

Finally, each frame had its procedures and actions related by enabling or causal relations which determined the dependencies between activities within a frame.The knowledge represented in Fig. 4.5 was derived from a task analysis of arranging a project meeting (task analysis is addressed in Chapters 11 and 12). The knowledge is assumed to be the knowledge utilized by the people who carried out this task.

Finally, each frame had its procedures and actions related by enabling or causal relations which determined the dependencies between activities within a frame.The knowledge represented in Fig. 4.5 was derived from a task analysis of arranging a project meeting (task analysis is addressed in Chapters 11 and 12). The knowledge is assumed to be the knowledge utilized by the people who carried out this task.

The purpose of the task analysis and knowledge representation was to assist in the design and evaluation of a messaging system to support a variety of messaging tasks.

The terms used in Fig. 4.5 were not technical, but were used to provide a common form of expression for all the different objects and actions that were encountered in the different instances of the task that were studied.

The purpose of the task analysis and knowledge representation was to assist in the design and evaluation of a messaging system to support a variety of messaging tasks.

The terms used in Fig. 4.5 were not technical, but were used to provide a common form of expression for all the different objects and actions that were encountered in the different instances of the task that were studied.

Therefore, the model is intended to be a generic model of the knowledge used to carry out the task.

For example, an information source could have been an address book, a diary, a note book or an online database. Similarly, an information token could have been a name, a telephone number, an address or a date.

The reasons for wanting such a generic model and the methods of producing them are considered in Chapters 11 and 12 on task analysis.

Therefore, the model is intended to be a generic model of the knowledge used to carry out the task.

For example, an information source could have been an address book, a diary, a note book or an online database. Similarly, an information token could have been a name, a telephone number, an address or a date.

The reasons for wanting such a generic model and the methods of producing them are considered in Chapters 11 and 12 on task analysis.

Scripts

Schank and Abelson (1977) developed a framework known as Scripts for repre-senting the specific examples of knowledge that people might have stored in memory.

A script is assumed to be a schema for frequently occurring sequences of events. For example, 'visits to a doctor', 'trips on a train' and the most frequently discussed, 'visits to a restaurant'.Scripts have variables (just as schemas and frames have).

Scripts

Schank and Abelson (1977) developed a framework known as Scripts for repre-senting the specific examples of knowledge that people might have stored in memory.

A script is assumed to be a schema for frequently occurring sequences of events. For example, 'visits to a doctor', 'trips on a train' and the most frequently discussed, 'visits to a restaurant'.Scripts have variables (just as schemas and frames have).

There are two categories of variable: roles and props. Roles are filled by persons and props are filled by objects. In addition to roles and props, scripts include a set of entry conditions which must prevail if the script is to be used (these being the context or scope of effect for the given script).

Scripts also include scenes and results. A scene is a particular grouping of activities within a script that can normally occur together and constitute a recognizable subset of the main activity.

There are two categories of variable: roles and props. Roles are filled by persons and props are filled by objects. In addition to roles and props, scripts include a set of entry conditions which must prevail if the script is to be used (these being the context or scope of effect for the given script).

Scripts also include scenes and results. A scene is a particular grouping of activities within a script that can normally occur together and constitute a recognizable subset of the main activity.

Consider the following scenario:Mary went to the restaurant. She ordered a quiche. Finally she paid the bill and then left.Now assume that the following script elements are associated with the above scenario:

Entry conditions hungry, had money, restaurant openRoles diner, waiter, cashierProps tables, money, chairs, menu, cutlery, food

Consider the following scenario:Mary went to the restaurant. She ordered a quiche. Finally she paid the bill and then left.Now assume that the following script elements are associated with the above scenario:

Entry conditions hungry, had money, restaurant openRoles diner, waiter, cashierProps tables, money, chairs, menu, cutlery, food

The script might then be as follows:

Entry scene Diner enters restaurant.Waiter seats diner at table.Waiter places menu on table.Diner begins to read menu.

Ordering scene Diner selects food from menu.Diner signals to waiter.Waiter approaches table.Diner orders food.Waiter leaves.

Eating scene Waiter brings food to table.Waiter leaves.Diner eats food with cutlery.Diner finishes eating food.

The script might then be as follows:

Entry scene Diner enters restaurant.Waiter seats diner at table.Waiter places menu on table.Diner begins to read menu.

Ordering scene Diner selects food from menu.Diner signals to waiter.Waiter approaches table.Diner orders food.Waiter leaves.

Eating scene Waiter brings food to table.Waiter leaves.Diner eats food with cutlery.Diner finishes eating food.

Leaving scene

Diner signals to waiter.Waiter approaches table.Diner asks waiter for bill.Waiter writes bill and gives to diner.

Diner checks bill.Diner approaches cashier.Diner gives cashier bill and money.

Cashier checks money.Diner leaves restaurant.

Leaving scene

Diner signals to waiter.Waiter approaches table.Diner asks waiter for bill.Waiter writes bill and gives to diner.

Diner checks bill.Diner approaches cashier.Diner gives cashier bill and money.

Cashier checks money.Diner leaves restaurant.

The script provides a structure for the temporal order of the elements of the activity and proivides sufficient information that can match the script to the instance of the activity of Mary going to a restaurant and having a quiche.Alongside scripts are plans.

Plans are more general and more abstract than scripts or schemas. Plans are formulated to satisfy goals and, as such, enable further actions to be initiated in an attempt to attain the goal. For example, consider the following scenario:

The script provides a structure for the temporal order of the elements of the activity and proivides sufficient information that can match the script to the instance of the activity of Mary going to a restaurant and having a quiche.Alongside scripts are plans.

Plans are more general and more abstract than scripts or schemas. Plans are formulated to satisfy goals and, as such, enable further actions to be initiated in an attempt to attain the goal. For example, consider the following scenario:

John knew that his wife's operation would be very expensive. There was always Uncle Harry. John reached for the area phone-book.

In order to understand this scenario, assume that John wants to borrow money from Uncle Harry and that he is reaching for the phone-book to find Uncle Harry's telephone number and intends to ring his uncle to ask for the money.There may not be a specific script for this event (unless borrowing money is a frequent activity). Consequently this activity is the result of some problem-solving behaviour.

First we can identify the problem as:

Problem = cost of operation

John knew that his wife's operation would be very expensive. There was always Uncle Harry. John reached for the area phone-book.

In order to understand this scenario, assume that John wants to borrow money from Uncle Harry and that he is reaching for the phone-book to find Uncle Harry's telephone number and intends to ring his uncle to ask for the money.There may not be a specific script for this event (unless borrowing money is a frequent activity). Consequently this activity is the result of some problem-solving behaviour.

First we can identify the problem as:

Problem = cost of operation

We then assume a structure of goals, subgoals and plans that might be as follows:

Primary goal Pay for operation.Plan Borrow money from Uncle Harry.Subgoal Contact Uncle Harry.Plan Call Uncle Harry on telephone.Subgoal Discover telephone number.Plan Look it up in telephone directory.

There would then be a detailed script for how to go about making a telephone call which would include how to look up numbers in the directory.Therefore, scripts are one form of knowledge structure for representing high-level aspects of knowledge and are capable of representing the temporal aspects of commonly occurring activities.

We then assume a structure of goals, subgoals and plans that might be as follows:

Primary goal Pay for operation.Plan Borrow money from Uncle Harry.Subgoal Contact Uncle Harry.Plan Call Uncle Harry on telephone.Subgoal Discover telephone number.Plan Look it up in telephone directory.

There would then be a detailed script for how to go about making a telephone call which would include how to look up numbers in the directory.Therefore, scripts are one form of knowledge structure for representing high-level aspects of knowledge and are capable of representing the temporal aspects of commonly occurring activities.

Analogical representationsAnalogical representations are assumed to be responsible for the representation of mental images.

The most common form of image is perhaps the visual image.

Visualization or visual imagery is assumed, often implicitly, to be partly responsible for the advantages that icon-based user interfaces might enjoy over textual user interfaces. The claim often made is that people find it easier to visualize than remember. This is clearly a misguided claim on a number of counts.

Analogical representationsAnalogical representations are assumed to be responsible for the representation of mental images.

The most common form of image is perhaps the visual image.

Visualization or visual imagery is assumed, often implicitly, to be partly responsible for the advantages that icon-based user interfaces might enjoy over textual user interfaces. The claim often made is that people find it easier to visualize than remember. This is clearly a misguided claim on a number of counts.

First, visualization is of course simply another form of remembering.

Second, some people find it very hard to visualize anything.

Third, most people find it very hard to visualize some things.

For example, it is easy to visualize a wastebasket if you have seen lots of examples of wastebaskets, since these are concrete objects which have a visible form.

First, visualization is of course simply another form of remembering.

Second, some people find it very hard to visualize anything.

Third, most people find it very hard to visualize some things.

For example, it is easy to visualize a wastebasket if you have seen lots of examples of wastebaskets, since these are concrete objects which have a visible form.

The ranges of wastebasket vary considerably and we would not expect everyone who reported that they could visualize a wastebasket to visualize the same sort.

Suppose, though, you were asked to visualize a more abstract or less concrete concept than a wastebasket, such as liberty or a data type.

These are two concepts which it would be hard for many people to visualize since they are not things which take a specific form.

The ranges of wastebasket vary considerably and we would not expect everyone who reported that they could visualize a wastebasket to visualize the same sort.

Suppose, though, you were asked to visualize a more abstract or less concrete concept than a wastebasket, such as liberty or a data type.

These are two concepts which it would be hard for many people to visualize since they are not things which take a specific form.

Most icon-based interfaces work well for concrete familiar objects and are always supported by a textual name which reinforces the user about the exact nature of the object being represented.

Interesting work on icons and imagery has been reported by Rogers (1986) and others.

Most icon-based interfaces work well for concrete familiar objects and are always supported by a textual name which reinforces the user about the exact nature of the object being represented.

Interesting work on icons and imagery has been reported by Rogers (1986) and others.

Rogers compared icons for concrete and abstract concepts and also icons with or without verbal labels.

She discovered that concrete icons with labels were the most readily understood and recognized form of presentation.

It is worth noting that there is also a relationship between the typicality of the item in the category and its usefulness as an icon.

Rogers compared icons for concrete and abstract concepts and also icons with or without verbal labels.

She discovered that concrete icons with labels were the most readily understood and recognized form of presentation.

It is worth noting that there is also a relationship between the typicality of the item in the category and its usefulness as an icon.

For example, a useful icon for a bird in most western cultures would be a robin or a sparrow, and not a hen or an ostrich, since robins and sparrows are more typical instances of birds than either ostriches or hens.

Similarly, an icon for an object should be a typical instance of the class of objects, rather than an obscure, atypical instance.

For example, a useful icon for a bird in most western cultures would be a robin or a sparrow, and not a hen or an ostrich, since robins and sparrows are more typical instances of birds than either ostriches or hens.

Similarly, an icon for an object should be a typical instance of the class of objects, rather than an obscure, atypical instance.

Imagery and its relation to icon design can be summarized along two dimensions.

First, there is the dimension of individual differences, which characterizes the extent to which people are capable of generating and thinking in terms of visual images.

Second, there is the dimension of the imagery potential of the concept; this characterizes the degree to which the concept itself has potential for being imagined.

Imagery and its relation to icon design can be summarized along two dimensions.

First, there is the dimension of individual differences, which characterizes the extent to which people are capable of generating and thinking in terms of visual images.

Second, there is the dimension of the imagery potential of the concept; this characterizes the degree to which the concept itself has potential for being imagined.

The individual differences between people in terms of their ability to generate mental images has been demonstrated using various types of tests and questionnaires.

Two categories of people's ability to generate images have been identified to be high vivid imagers and low vivid imagers.

The individual differences between people in terms of their ability to generate mental images has been demonstrated using various types of tests and questionnaires.

Two categories of people's ability to generate images have been identified to be high vivid imagers and low vivid imagers.

High vivid imagers report that they can easily generate visual images while low vivid imagers report (and the tests associated with these reports confirm this) that they find it difficult or impossible to generate visual images.

The second dimension of imagery is the imagery potential of the concept.

High vivid imagers report that they can easily generate visual images while low vivid imagers report (and the tests associated with these reports confirm this) that they find it difficult or impossible to generate visual images.

The second dimension of imagery is the imagery potential of the concept.

Extensive research has shown that, even amongst people who are rated as being high vivid imagers, there are some concepts which they find it difficult to imagine.

The results of this research show that concrete concepts can be imagined far more readily than abstract concepts.

Extensive research has shown that, even amongst people who are rated as being high vivid imagers, there are some concepts which they find it difficult to imagine.

The results of this research show that concrete concepts can be imagined far more readily than abstract concepts.

It appears then that visual imagery is something which is not necessarily universal across all people and not all concepts are capable of being imagined.

Consequently, icons rely in part on their association with a concept and the person's image of that concept.

It appears then that visual imagery is something which is not necessarily universal across all people and not all concepts are capable of being imagined.

Consequently, icons rely in part on their association with a concept and the person's image of that concept.

Icons, therefore, must be used carefully since not all people can easily form images, and those that can cannot easily imagine abstract concepts. Bell (1989) found that icons without labels are not significantly more effective than labels.

Furthermore, for some concepts, labels alone are better than icons alone.

Icons, therefore, must be used carefully since not all people can easily form images, and those that can cannot easily imagine abstract concepts. Bell (1989) found that icons without labels are not significantly more effective than labels.

Furthermore, for some concepts, labels alone are better than icons alone.

There are of course nonvisual images which might include tastes, smells, sounds and perhaps even feelings. However, very little work has yet been carried out in HCI to understand the utility of these kinds of images.As a demonstration of imagery, imagine the scene described by the following passage:

There are of course nonvisual images which might include tastes, smells, sounds and perhaps even feelings. However, very little work has yet been carried out in HCI to understand the utility of these kinds of images.As a demonstration of imagery, imagine the scene described by the following passage:

As a demonstration of imagery, imagine the scene described by the following passage:

It was a hot summer's day. The sky was clear and blue. The sun glistened on the lake and the branches of the trees were reflected in the water. The couple were sat on the grass, under the willow tree, above a small pebble beach. Children were playing in the water nearby. Out on the lake there was a number of sailing boats.

As a demonstration of imagery, imagine the scene described by the following passage:

It was a hot summer's day. The sky was clear and blue. The sun glistened on the lake and the branches of the trees were reflected in the water. The couple were sat on the grass, under the willow tree, above a small pebble beach. Children were playing in the water nearby. Out on the lake there was a number of sailing boats.

One was gliding slowly towards them. It was the most magnificent boat on the lake with a gleaming white hull and a tall mast with three sails. The couple could hardly believe that just a few hours before they had left the heat and grime of the city. It seemed as if the day had been made for them.

What colour were the sails of the boat?

Most people readily give an answer to this question and often add that it was the colour of the sail in their mental image.

Clearly, it was not in the story. This is a rather simple example of imagery.

Most people readily give an answer to this question and often add that it was the colour of the sail in their mental image.

Clearly, it was not in the story. This is a rather simple example of imagery.

More interesting experimental investigations of mental imagery have suggested that the objects in an image can be subjected to transformations.

This has been the basis for some researchers (e.g. Shepard, 1978, Paivio, 1978) to suggest that the knowledge underlying images may be analogical rather than propositional.

More interesting experimental investigations of mental imagery have suggested that the objects in an image can be subjected to transformations.

This has been the basis for some researchers (e.g. Shepard, 1978, Paivio, 1978) to suggest that the knowledge underlying images may be analogical rather than propositional.

Shepard and his colleagues have shown in numerous experiments that subjects appear to be mentally rotating or otherwise transforming an object (see, for example, Shepard, 1978).

He argues that the process of mentally rotating an object involves a mental analogue of a physical rotation.

Shepard and his colleagues have shown in numerous experiments that subjects appear to be mentally rotating or otherwise transforming an object (see, for example, Shepard, 1978).

He argues that the process of mentally rotating an object involves a mental analogue of a physical rotation.

It is assumed that the representation is being processed in an analogous manner to that which would go on if the subject were actually perceiving an external object physically rotating.

Furthermore, it is suggested that the internal representation passes through a certain trajectory of intermediate steps, each of which has a one-to-one correspondence with a physical rotation of the imaged object.

It is assumed that the representation is being processed in an analogous manner to that which would go on if the subject were actually perceiving an external object physically rotating.

Furthermore, it is suggested that the internal representation passes through a certain trajectory of intermediate steps, each of which has a one-to-one correspondence with a physical rotation of the imaged object.

The classic paradigm for Shepard's experiments involves presenting subjects with a complex object and then later presenting them with a set of several objects which includes the original object rotated (in one or more planes).

The results of these studies generally show that as the angle of rotation increases, the time it takes for the subject to identify the object as being in the set also increases.

The classic paradigm for Shepard's experiments involves presenting subjects with a complex object and then later presenting them with a set of several objects which includes the original object rotated (in one or more planes).

The results of these studies generally show that as the angle of rotation increases, the time it takes for the subject to identify the object as being in the set also increases.

From these studies it is assumed that the representation of the imagined object is analogous, in that it preserves some degree of spatial structuring of the external object.

However, it should be recognized that we could produce a propositional representation of spatial properties (for example, next to, adjoining, above, below, etc.).

From these studies it is assumed that the representation of the imagined object is analogous, in that it preserves some degree of spatial structuring of the external object.

However, it should be recognized that we could produce a propositional representation of spatial properties (for example, next to, adjoining, above, below, etc.).

Work by Kosslyn (see, for example, Kosslyn, 1980) has lead to the formulation of a theory of image representation which is based on the analogy of a cathode ray tube (CR1) and comprises two layers of representation.

The first layer is a surface representation which has the following properties:

Work by Kosslyn (see, for example, Kosslyn, 1980) has lead to the formulation of a theory of image representation which is based on the analogy of a cathode ray tube (CR1) and comprises two layers of representation.

The first layer is a surface representation which has the following properties:

1 Part of the image represents corresponding parts of the object, preserving such properties as distance between parts of the object.

2 There is a limited spatial extent, i.e. images can overflow if too large.

3 The surface representation has a grain size and detail is lost if the image is too small.

4 There is periodic refreshing of the image otherwise fading occurs.

1 Part of the image represents corresponding parts of the object, preserving such properties as distance between parts of the object.

2 There is a limited spatial extent, i.e. images can overflow if too large.

3 The surface representation has a grain size and detail is lost if the image is too small.

4 There is periodic refreshing of the image otherwise fading occurs.

The second layer is a deep representation which generates the image from a propositional representation in long-term memory.

A third component of the theory is a 'mind's eye' which interfaces between the surface image and the deep representation, and which uses parts of the visual system to process or interpret the image.

This processor performs functions such as 'generates' the image, 'looks for' parts of the image, 'transforms' the image by 'scan', 'zoom'. 'pan' and 'rotate' subprocessors.

The second layer is a deep representation which generates the image from a propositional representation in long-term memory.

A third component of the theory is a 'mind's eye' which interfaces between the surface image and the deep representation, and which uses parts of the visual system to process or interpret the image.

This processor performs functions such as 'generates' the image, 'looks for' parts of the image, 'transforms' the image by 'scan', 'zoom'. 'pan' and 'rotate' subprocessors.

It is best not to think of imagery as a separate representational system but as a function of some special-purpose processor.

Johnson (1982), in a series of experiments, has shown that subjects can generate images of simple linear arm movements which are functionally equivalent to actual arm movements.

It is best not to think of imagery as a separate representational system but as a function of some special-purpose processor.

Johnson (1982), in a series of experiments, has shown that subjects can generate images of simple linear arm movements which are functionally equivalent to actual arm movements.

The functional equivalence is in terms of the spatial coordinates for the movement and the image.

The parameters of the movement and the image are assumed to be one and the same and are abstract representations of spatial properties (such as starting-point, direction, distance, end point).

The functional equivalence is in terms of the spatial coordinates for the movement and the image.

The parameters of the movement and the image are assumed to be one and the same and are abstract representations of spatial properties (such as starting-point, direction, distance, end point).

Procedural representations

We can distinguish between knowledge about something (i.e. factual or declarative knowledge) and knowledge about how to do something (i.e. procedural knowledge).

Procedural representations

We can distinguish between knowledge about something (i.e. factual or declarative knowledge) and knowledge about how to do something (i.e. procedural knowledge).

For example, declarative knowledge of a bicycle would include knowledge of its parts (pedals, saddle, frame, wheels, handlebars, etc.), what it could be used for, and many other items of knowledge, such as examples of good bicycles, where to purchase one, etc.

In contrast, procedural knowledge associated with a bicycle might include how to ride and how to repair one.

For example, declarative knowledge of a bicycle would include knowledge of its parts (pedals, saddle, frame, wheels, handlebars, etc.), what it could be used for, and many other items of knowledge, such as examples of good bicycles, where to purchase one, etc.

In contrast, procedural knowledge associated with a bicycle might include how to ride and how to repair one.

There is assumed to be a relationship between the type of knowledge (procedural or declarative) and consciousness.

It is assumed that declarative knowledge is accessible to consciousness in that it can be examined and combined, while procedural knowledge is thought to be inaccessible to consciousness.

There is assumed to be a relationship between the type of knowledge (procedural or declarative) and consciousness.

It is assumed that declarative knowledge is accessible to consciousness in that it can be examined and combined, while procedural knowledge is thought to be inaccessible to consciousness.

It is also generally assumed that the only way to access the procedural knowledge of riding a bicycle is to engage in the activity of riding a bicycle.

Note that it is not just physical actions that are represented in procedures.

For example, knowledge of how to perform arithmetical calculations will also be represented by procedural knowledge.

It is also generally assumed that the only way to access the procedural knowledge of riding a bicycle is to engage in the activity of riding a bicycle.

Note that it is not just physical actions that are represented in procedures.

For example, knowledge of how to perform arithmetical calculations will also be represented by procedural knowledge.

Knowledge of this form is assumed to include representations of action procedures, and these procedures are tailored for the performance of specific actions.

In contrast, declarative knowledge (represented as propositions) can be used for a variety of purposes.

A further distinction between procedural and declarative knowledge is in the context of skilled performance.

Knowledge of this form is assumed to include representations of action procedures, and these procedures are tailored for the performance of specific actions.

In contrast, declarative knowledge (represented as propositions) can be used for a variety of purposes.

A further distinction between procedural and declarative knowledge is in the context of skilled performance.

One characteristic of skilled perform-ance is that the performer has developed highly tuned procedures that support the performance and allow it to be executed with ease and efficiency of effort.

In contrast, unskilled performance is assumed to lack the highly developed procedural knowledge that is optimum for the task, and as a consequence the performance is difficult and inefficient.

One characteristic of skilled perform-ance is that the performer has developed highly tuned procedures that support the performance and allow it to be executed with ease and efficiency of effort.

In contrast, unskilled performance is assumed to lack the highly developed procedural knowledge that is optimum for the task, and as a consequence the performance is difficult and inefficient.

One concern about procedural representations is how they might be activated.

One possibility is that a procedure is activated by direct invocation, that is, some other procedure or an interpreter calls the required procedure.

Alternatively, a triggering mechanism might be integrated into the procedure. In this case the procedure would monitor a database for relevant data structures.

One concern about procedural representations is how they might be activated.

One possibility is that a procedure is activated by direct invocation, that is, some other procedure or an interpreter calls the required procedure.

Alternatively, a triggering mechanism might be integrated into the procedure. In this case the procedure would monitor a database for relevant data structures.

When a relevant data structure had been identified, the procedure would be fired. However, for this triggering mechanism to work there must be some form of demon or process which monitors the database.

One solution to the representation of procedures is found in the form of production rules and production systems (PS).

When a relevant data structure had been identified, the procedure would be fired. However, for this triggering mechanism to work there must be some form of demon or process which monitors the database.

One solution to the representation of procedures is found in the form of production rules and production systems (PS).

Production rules include a demon or process which monitors and fires the action. A production rule consists of

‘if ‘ ’then'or‘condition' 'action'statements,

IF (condition/triggering) THEN (do these actions)

Production rules include a demon or process which monitors and fires the action. A production rule consists of

‘if ‘ ’then'or‘condition' 'action'statements,

IF (condition/triggering) THEN (do these actions)

Production systems are collections of connected production rules and as such constitute a complete knowledge structure for a given activity.

PSs are modular in format but new production rules can be added to existing PSs or alternatively PSs can be replaced by new, more powerful PSs.

Production systems are collections of connected production rules and as such constitute a complete knowledge structure for a given activity.

PSs are modular in format but new production rules can be added to existing PSs or alternatively PSs can be replaced by new, more powerful PSs.

In this way, learning and improvements in performance can occur.

However, that requires us to consider how people acquire skills.

Skill and skill acquisition are considered in Chapter 5.

In this way, learning and improvements in performance can occur.

However, that requires us to consider how people acquire skills.

Skill and skill acquisition are considered in Chapter 5.

CONCLUSION

Representations can be thought of as notations.

Therefore, the problem of identifying the form of knowledge representation can be projected as a set of constraints on a notation.

CONCLUSION

Representations can be thought of as notations.

Therefore, the problem of identifying the form of knowledge representation can be projected as a set of constraints on a notation.

First, any notation for knowledge

representation should be rich enough to

represent all of the relevant knowledge

structures and cognitive processes that

might act on them.

First, any notation for knowledge

representation should be rich enough to

represent all of the relevant knowledge

structures and cognitive processes that

might act on them.

Second, those processes that are

assumed to be easily carried out should

in fact be easily carried out through the

chosen representation.

Second, those processes that are

assumed to be easily carried out should

in fact be easily carried out through the

chosen representation.

Representations consist of two parts, the data structures that are stored according to the chosen representational format, and the processes that are able to operate on the data structures.

Representations consist of two parts, the data structures that are stored according to the chosen representational format, and the processes that are able to operate on the data structures.

The knowledge that people recruit and

use to carry out tasks will include each

type of knowledge, procedural,

declarative and analogical.

The knowledge that people recruit and

use to carry out tasks will include each

type of knowledge, procedural,

declarative and analogical.

There is no reason to believe that each of these types of knowledge is discrete and used exclusively in certain tasks.

Instead, we should think of these as functional descriptions of the different types of knowledge that people use.

There is no reason to believe that each of these types of knowledge is discrete and used exclusively in certain tasks.

Instead, we should think of these as functional descriptions of the different types of knowledge that people use.

The representational differences that have been postulated in psychological research are less important than the functional differences between the differing types of knowledge.

The representational differences that have been postulated in psychological research are less important than the functional differences between the differing types of knowledge.

For example, it seems to be less important to consider whether images are generated from an analogue or propositional representation than to consider the functional significance of images in the role of remembering, understanding and thinking.

For example, it seems to be less important to consider whether images are generated from an analogue or propositional representation than to consider the functional significance of images in the role of remembering, understanding and thinking.

Our main concerns in the next chapter is how we use our knowledge and how we acquire skill and expertise.

The different types of representation discussed here for propositional and procedural knowledge form a part of a theory of skill acquisition that describes how a skill is acquired and how we develop more powerful (appropriate) procedural knowledge.

Our main concerns in the next chapter is how we use our knowledge and how we acquire skill and expertise.

The different types of representation discussed here for propositional and procedural knowledge form a part of a theory of skill acquisition that describes how a skill is acquired and how we develop more powerful (appropriate) procedural knowledge.

Exercises

1. Distinguish between represented and representing 'worlds'.

2. Construct a representation of the knowledge you assume a person must possess to carry out a simple task such as creating a new document file. Remember to represent both the factual and event knowledge.

Exercises

1. Distinguish between represented and representing 'worlds'.

2. Construct a representation of the knowledge you assume a person must possess to carry out a simple task such as creating a new document file. Remember to represent both the factual and event knowledge.

Exercises

3 Design and carry out a simple experiment to investigate the use of concrete and abstract icons, with and without labels, for the ease with which people can understand their meaning, and remember their function.

4 Produce a propositional knowledge representation using production rules for the knowledge required to carry out a simple task using any application program you choose.

Exercises

3 Design and carry out a simple experiment to investigate the use of concrete and abstract icons, with and without labels, for the ease with which people can understand their meaning, and remember their function.

4 Produce a propositional knowledge representation using production rules for the knowledge required to carry out a simple task using any application program you choose.